Are data centers the new REIT? Not quite — but Meta’s new mega–data center in northeast Louisiana marks what one expert calls a “decisive shift” in how hyperscalers finance the AI era: by turning data centers into a new investable asset class.
“This is where capital markets meet compute,” said Sean McDevitt, a partner at management consulting firm Arthur D. Little, which provided commercial due diligence advice to
Meta.
Traditionally, tech giants like Meta,
Google, and
Microsoft have funded their data center buildouts directly. This time, Meta is partnering with Blue Owl Capital, a private-credit investment firm, on the $27 billion data center known as Hyperion. As reported by
The Wall Street Journal, Blue Owl owns 80% of the project, while Meta holds 20%, operating and leasing the facility long-term.
BlackRock bought more than $3 billion of bonds that the joint venture (dubbed Beignet) issued last week to finance the project, in a sale arranged by
Morgan Stanley.
The deal stands out for its scale—the largest private-debt offering ever—and for its A+ rating from S&P, which reflects Meta’s backing of the project (albeit with just a single agency rating). Yet the debt had a yield of 6.58% at issue, a level closer to high-yield, or “junk,” bond territory.
That structure allows Meta to build its data center without putting the full $27 billion of debt on its own balance sheet. The approach—known as a special-purpose vehicle (SPV) or off–balance-sheet financing—is largely new territory for hyperscale infrastructure.
“By being able to access outside capital, you’re not limited to your own free cash flow generation,” McDevitt said. “You’re bringing on investors with return profiles on an infrastructure-type investment that allows companies to build bigger, larger, quicker, and faster.” He compared it to taking out a mortgage: you can buy a bigger house—or, in this case, build more data centers—by borrowing instead of paying cash up front.
McDevitt believes the Hyperion deal could become a template for the industry. He estimates that roughly $150 billion in AI-driven data center construction is coming in the next few years. If other hyperscalers—Microsoft, Google,
Amazon, and OpenAI among them—adopt similar models, capital markets rather than tech companies themselves will effectively fund the infrastructure of the AI era. “This is replicable,” he said, though he cautioned that it remains to be seen how the project performs in practice.
“Now what has to happen? Meta has to build this thing, then put workloads in it and operate under the presumption that they’ll monetize those computing loads driven by AI in the future,” McDevitt added.
That is precisely where criticism of the deal lies: According to Global Data Center Hub
analysis, “If AI workloads or margins stumble, these SPVs could echo the dark-fiber overbuild of the 1990s vast capacity sitting idle while debt remains outstanding.”
Still, for now, there’s no reason to think other major banks won’t try the same thing, said McDevitt. “Why wouldn’t others look to mimic [this deal]?”
Sharon Goldman
X: @sharongoldmanEmail:
sharon.goldman@fortune.comSubmit a deal for the Term Sheet newsletter
here.
Joey Abrams curated the deals section of today’s newsletter. Subscribe here.
